al.
10.1093/mnras/stv1379
DOI
http://hdl.handle.net/20.500.12386/26270
Handle
MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY
Journal
452
Number
MNRAS 452, 1112–1127 (2015) doi:10.1093/mnras/stv1379
Periodic signals from the Circinus region: two new cataclysmic variables
and the ultraluminous X-ray source candidate GC X-1
P. Esposito,
1,2‹G. L. Israel,
3D. Milisavljevic,
2M. Mapelli,
4,5L. Zampieri,
4L. Sidoli,
1G. Fabbiano
2and G. A. Rodr´ıguez Castillo
31INAF–Istituto di Astrofisica Spaziale e Fisica Cosmica – Milano, via E. Bassini 15, I-20133 Milano, Italy 2Harvard–Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA
3INAF–Osservatorio Astronomico di Roma, via Frascati 33, I-00040 Monteporzio Catone, Italy 4INAF–Osservatorio Astronomico di Padova, vicolo dell’Osservatorio 5, I-35122 Padova, Italy 5INFN, Sezione di Milano Bicocca, piazza della Scienza 3, I-20126 Milano, Italy
Accepted 2015 June 18. Received 2015 June 9; in original form 2015 May 14
A B S T R A C T
The examination of two 2010 Chandra ACIS (Advanced CCD Imaging Spectrom-eter) exposures of the Circinus galaxy resulted in the discovery of two pulsators: CXO J141430.1−651621 and CXOU J141332.9−651756. We also detected 26 ks pulsations in CG X-1, consistently with previous measures. For∼40 other sources, we obtained limits on periodic modulations. In CXO J141430.1−651621, which is ∼2 arcmin outside the Circinus galaxy, we detected signals at 6120± 1 s and 64.2 ± 0.5 ks. In the longest observation, the source showed a flux of≈1.1 × 10−13erg cm−2s−1(absorbed, 0.5–10 keV) and the spectrum could be described by a power law with photon index 1.4. From archival observations, we found that the luminosity is variable by≈50 per cent on time-scales of weeks to years. The two periodicities pin down CXO J141430.1−651621 as a cataclysmic variable of the inter-mediate polar subtype. The period of CXOU J141332.9−651756 is 6378 ± 3 s. It is located inside the Circinus galaxy, but the low absorption indicates a Galactic foreground object. The flux was≈5 × 10−14erg cm−2 s−1 in the Chandra observations and showed≈50 per cent variations on weekly/yearly scales; the spectrum is well fitted by a power law with 0.9. These characteristics and the large modulation suggest that CXOU J141332.9−651756 is a magnetic cataclysmic variable, probably a polar. For CG X-1, we show that if the source is in the Circinus galaxy, its properties are consistent with a Wolf–Rayet (WR) plus black hole (BH) binary. We consider the implications of this for ultraluminous X-ray sources and the prospects of Advanced LIGO and Virgo. In particular, from the current sample of WR–BH systems, we estimate an upper limit to the detection rate of stellar BH–BH mergers of∼16 yr−1.
Key words: novae, cataclysmic variables – galaxies: individual: Circinus – X-rays:
bi-naries – X-rays: individual: CG X-1 (CXOU J141312.3−652013) – X-rays: individual: CXOU J141332.9−651756 – X-rays: individual: CXO J141430.1−651621.
1 I N T R O D U C T I O N
The Chandra ACIS Timing Survey at Brera And Rome astronom-ical observatories project (CATS @ BAR; Israel et al., in prepara-tion) is a Fourier-transform-based systematic search for new pulsat-ing sources in the Chandra Advanced CCD Imagpulsat-ing Spectrometer (ACIS; Garmire et al.2003) public archive. As of 2015 April 30, 10 282 ACIS timed exposure observations have been examined and∼457 000 sources were detected. Data taken with gratings or
E-mail:paoloesp@iasf-milano.inaf.it
in continuous-clocking mode were not considered. The∼93 600 light curves of sources with more than 150 photons were searched for coherent signals with an algorithm based on that of Israel & Stella (1996). The limit of 150 counts is related to the intrinsic ability of the Fourier transform to detect a signal with 100 per cent modulation at a minimum confidence level of 3.5σ in 105–106trials.
CATS @ BAR has so far discovered 43 new certain X-ray pulsators; see Esposito et al. (2013a,b,c) for the first results and Esposito et al. (2014,2015) for our analogous Swift project.
In this paper, we report on the CATS @ BAR results for the galaxy ESO 97-G13 (the ‘Circinus galaxy’, hereafter CG; Freeman et al.1977) and its surroundings in the Circinus constellation. CG
2015 The Authors
Chandra ACIS-235678 356 2000 Mar. 14 24.7 TE FAINT (3.24 s)
Chandra ACIS-235678 2454 2001 May 02 4.4 TE FAINT (3.24 s)
XMM* pn/MOS 1/MOS 2 0111240101 2001 Aug. 6–7 100.6/104.1/104.0 FF (73.4 ms)/FF (2.6 s)/LW (0.9/2.7 s)
Chandra ACIS-456789 10873 2009 Mar. 01 18.1 TE HETG VFAINT (2.04 s)
Chandra ACIS-456789 10850 2009 Mar. 03 13.8 TE HETG VFAINT (2.04 s)
Chandra ACIS-456789 10872 2009 Mar. 04 16.5 TE HETG VFAINT (2.04 s)
Chandra# ACIS-23678 12823 2010 Dec. 17–19 154.4 TE VFAINT (3.14 s)
Chandra# ACIS-23678 12824 2010 Dec. 24 39.4 TE VFAINT (3.14 s)
XMM pn/MOS 1/MOS 2 0656580601 2014 Mar. 01 32.7/37.8/37.9 FF (73.4 ms)/FF (2.6 s)/LW (2.7 s)
Notes.aTE: timed exposure, HETG: high-energy transmission grating, VFAINT: very faint telemetry format, FF: full frame, LW: large window; the readout time is given in parentheses, for the central and peripheral CCDs in the case of the MOS 2 in LW.
is a nearby Seyfert II active galaxy that lies close to the plane of our own Galaxy (J2000 Galactic coordinates: l= 311.◦3, b= −03.◦8; distance d= 4.2 Mpc; Tully et al.2009). The CG contains hydrogen-rich star-forming regions in the inner spiral arms and, due to its closeness, offers a good opportunity to study its population of X-ray sources (Bauer et al.2001; Sambruna et al.2001), which includes several ultraluminous X-ray sources (ULXs; Winter, Mushotzky & Reynolds2006; Mapelli et al.2010a; Walton et al.2013; see Mushotzky2004; Fabbiano2006; Zampieri & Roberts2009; Feng & Soria2011for reviews on ULXs).
The CG was observed several times with Chandra, but it was in two long timed exposure observations carried out in late 2010 that CATS @ BAR pinpointed two new X-ray pulsators in the Circinus region: CXO J141430 and the uncatalogued CXOU J141332. The pipeline detected also CG X-1 (CXOU J141312.3−652013), whose emission is modulated at∼7 h. CG X-1 is not a new pulsator, but the longstanding debate about its nature (Bauer et al.2001; Smith & Wilson2001; Weisskopf et al.2004) prompted us to include it in our study.
The plan of the paper is as follows. In Section 2, we give details on the X-ray observations used in our study. The rest of the paper is divided into two main parts. The first one focuses on the new pulsators and comprises Sections 3 to 7. In Section 3, we describe the timing analysis that led to the discovery of the new pulsators and allowed us also to set upper limits on the pulsations for dozens of other X-ray sources. The detailed study of CXO J141430 is pre-sented in Section 4 and that of CXOU J141332 in Section 5. To study these sources, we also used data from XMM–Newton and op-tical observations taken with the VLT Survey Telescope (VST). The optical observations and their analysis are described in Section 6. The nature of CXO J141430 and CXOU J141332 is discussed in Section 7.
The second part of the paper is dedicated to CG X-1. In Section 8, we recall the main facts about this source. The results from the analysis of the 2010 Chandra data, which were not used before to study CG X-1, are summarized in Section 9. In Section 10, we propose that CG X-1 might be a Wolf–Rayet plus black hole (WR– BH) binary system, and consider the implications of this possibility for ULXs and for the prospects of detection of gravitational radiation from BH–BH mergers. A summary with conclusive remarks follows in Section 11.
2 X - R AY O B S E RVAT I O N S
All the observations used in this work are summarized in Ta-ble 1. The most important observations are the ones in which CATS @ BAR detected the new pulsators and its companion, 12823 and 12824, marked with a hash mark in Table1. They were carried out in a week in 2010 December to study the central region of the CG (Mingo et al.2012). Their combined exposure is∼190 ks. In both cases, three ACIS-S and two ACIS-I CCDs were used in full frame mode, ensuring a wide coverage over the Circinus region. The Chandra data were processed and analysed with the Chandra Inter-active Analysis of Observations (CIAO) software package (version 4.7; Fruscione et al.2006) and the calibration files inCALDBversion
4.6.7. The Circinus field as imaged with Chandra in observation 12823 is shown in Fig.1. In the data sets 12823 and 12824, CG X-1 and CXOU J141332 were positioned in the back-illuminated CCD 7 (S3). The photons from these sources were accumulated within a circle with radius 1.5 arcsec and an ellipse with semi-axes 3.5 and 3 arcsec, respectively. CXO J141430 fell on the front-illuminated CCD 3 (I3) and the source counts were extracted from an ellipse with semi-axes of 15 and 14 arcsec. The choice of regions of dif-ferent sizes is due to the point spread function at the off-axis angles of the sources. For each source, the background was estimated lo-cally, using source-free regions as close as possible to the target. The Solar system barycentre correction to the photon arrival times was applied withAXBARY. The spectra, the redistribution matrices,
and the ancillary response files were created usingSPECEXTRACT.
The second most useful observation for our study is that per-formed with XMM–Newton in 2001 August with a duration of ∼100 ks (obs. ID 0111240101; Molendi, Bianchi & Matt2003; it is marked by a star in Table1). We used the data collected with the European Photon Imaging Camera (EPIC), which consists of two MOS (Turner et al.2001) and one pn (Str¨uder et al.2001) CCD detectors. The raw data were reprocessed using the XMM–Newton Science Analysis Software (SAS, version 14.0) and the calibration
files in the CCF release of 2015 March. The observation suffered intense soft-proton flares. The intervals of flaring background were located by intensity filters (see e.g. De Luca & Molendi2004) and excluded from the analysis. This reduced the net exposure time by∼30 per cent in the pn back-illuminated CCDs and ∼10 per cent in the MOS front-illuminated CCDs. The source photons were extracted from circles with radius of 25 arcsec for CXO J141430
1114
P. Esposito et al.
Figure 1. Chandra/ACIS image of the Circinus field. North is top and east is left; each CCD subtends an 8.4 arcmin× 8.4 arcmin square on the sky and the
zoomed area is 1.2 arcmin× 1 arcmin. Circles mark the detected sources. Red circles indicate sources with >150 counts for which we performed a timing analysis and blue circles the sources from which periodic signals were detected. The labelled sources (with the ACIS CCD number and an identification number) are those for which either signals were detected (source 7–3 is CG X-1, 7–20 is CXOU J141332 and 3–1 is CXO J141430) or an upper limit on the pulsed fraction could be placed (see Fig.2). The solid and dashed ellipses indicate the size of the CG from 90 per cent total B light and total infrared (2MASS) magnitude, respectively (from the NASA/IPAC Extragalactic Database, NED, seehttp://ned.ipac.caltech.edu/). Source 7–1 is the CG’s active galactic nucleus (AGN), 7–2 is the young supernova remnant candidate in the CG (CG X-2; Bauer et al.2001); other notable sources are the ultraluminous X-ray sources 7–5= Circinus ULX3, 7–16 = Circinus ULX4, 7–4 = Circinus XMM1 = ULX5, 7–17 = Circinus XMM2 [while Circinus XMM3 is undetected, we used the nomenclature of Winter et al. (2006), Gladstone et al. (2013), and Walton et al. (2013)].
and 15 arcsec for CXOU J141332 (these radii were essentially im-posed by CCD gaps and/or the presence of neighbouring sources) and the backgrounds from regions in the same chip as the sources. CXOU J141332 was in the unread part of the central CCD of the MOS 2 operated in a partial window mode, so only pn and MOS 1 data exist for it. Photon arrival times were converted to the Solar system barycentre using the SAS taskBARYCEN. The ancillary
re-sponse files and the spectral redistribution matrices for the spectral analysis were generated withARFGENandRMFGEN, respectively. Due
to the low number of photons, we combined for each source the
spectra from the available EPIC cameras and averaged the response files usingEPICSPECCOMBINE.
We made use of other Chandra and XMM–Newton observations, which were reduced and analysed in a similar way; they provided only detections and flux estimates, or upper limits for the two new pulsators. Six data sets were collected with Chandra from 2000 to 2009 with various instrumental setups and durations from∼1 to 25 ks, and one with XMM–Newton in 2014 with exposure of∼30 ks. Apart from these observations, listed in Table1, no other Chandra pointing of the CG covered the positions of CXO J141430 or
Figure 2. Upper limits on the pulsed fraction of the Circinus sources. Red circles indicate sources detected in the back-illuminated ACIS-7 (S3) CCD and blue squares sources from front-illuminated CCDs. The sources are identified by the same labels as in Fig.1. Dashed vertical lines mark approximate fluxes for the S3 CCD sources. Source 7–1 is the CG’s AGN; other notable sources are: 7–5= Circinus ULX3, 7–16 = Circinus ULX4, and 7–4 = Circinus XMM1 = ULX5.
CXOU J141332, while in an∼60 ks observation performed with XMM–Newton in 2013 (obs. ID 0701981001), both sources fell ei-ther in gaps or at the edge of CCDs, or outside the field of view of the instruments.
3 C AT S @ B A R T I M I N G A N A LY S I S
TheCIAO WAVEDETECTroutine detected 156 sources in the ACIS field
of view of observation 12823; they are marked by circles in Fig.1. The 44 sources with more than 150 photons (those marked by ma-genta and red circles in Fig.1) were searched for periodic signals. The CATS @ BAR search algorithm is based on a fast Fourier transform and takes into account also the possible presence of addi-tional non-Poissonian noise components in the Leahy-normalized (Leahy et al.1983) power spectra (see Israel & Stella1996for more details). Correspondingly, the CATS @ BAR signal threshold in a power spectrum, which takes into account the number of indepen-dent Fourier frequencies, is also a function of the local underlying noise. For the Circinus data, the maximum frequency of the search (∼0.16 Hz) is dictated by the sampling time of 3.14 s, while the longest period to which the observation is realistically sensitive (be-cause of its duration) is≈80 ks; 32 768 frequencies were searched. The search resulted in the detection of five sources with significant signals in their power spectra.
In two cases, the power peaks were coincident with the frequen-cies of known spurious signals due to the spacecraft dithering pat-tern. The CATS @ BAR pipeline automatically performs check for these artificial signals by means of theCIAOtaskDITHER_REGION.1
Furthermore, every candidate signal is cross-checked with the
1Seehttp://cxc.harvard.edu/ciao/ahelp/dither_region.html.
CATS @ BAR data base of recurring signals of instrumental ori-gin, and repeating or dubious signals are carefully inspected and rejected. These two sources are the supernova remnant candidate CG X-2 (Bauer et al.2001), labelled 7–2 in Fig.1, and Circinus XMM2, which is classified as an ULX (Winter et al. 2006), la-belled 7–17. In a third object, CG X-1, the detected∼27-ks-period modulation was already known (Bauer et al.2001; object 7–3 in Fig.1). This source is discussed in detail in Sections 8–10. The remaining two sources, 3–1= CXO J141430 (Section 4) and the uncatalogued 7–20= CXOU J141332 (Section 5), are genuine new X-ray pulsators, as was also confirmed by the other data sets.
For all the other sources with more than 150 events, a 3σ upper limit to the pulsed fraction of any sinusoidal signal was calcu-lated (throughout the paper, we will give 3σ upper limits on non-detections and limits at the 90 per cent confidence level on poorly constrained quantities; all uncertainties will be given at the 1σ confi-dence level). The pulsed fraction was defined as the semi-amplitude of the sinusoidal modulation divided by the mean count rate. As expected, around 100–200 photons the upper limits start crossing the 100 per cent threshold, above which no meaningful informa-tion related to any coherent signal can be inferred. For many other sources, the upper limits are not constraining. For future reference, all the results are summarized in Fig.2.
4 T H E 1 . 7 / 1 7 . 8 H P U L S AT O R : C X O J 1 4 1 4 3 0 4.1 Timing analysis
CXO J141430 is the brightest of the two new CATS @ BAR pul-sators. It shows two distinct periodic signals: an∼100 per cent mod-ulation at about 6.1 ks and another large-amplitude signal at about 64 ks. The power spectrum is shown in Fig.3. When the 32 768
1116
P. Esposito et al.
Figure 3. Left-hand panel: 0.5–10 keV power spectra of CXO J141430 obtained from Chandra/ACIS (top, black line, observations 12823 and 12824 combined) and XMM–Newton/EPIC (bottom, red line) data. For displaying purposes, the Chandra data have been shifted in power by+100. The corresponding Chandra (top) and XMM–Newton (bottom) light curves folded to the best periods are shown in the panels on the right for P1= 6.1 ks. In the inset is shown the Chandra light curve folded to the longer second period P2= 64.2 ks.
frequencies analysed and the number of sources for which the search was carried out (44) are taken into account, both signals were de-tected at a confidence level larger than 10σ .
To hone the estimates of the periods, we made use of both the Chandra 12823 and 12824 pointings, where about 820 and 250 net counts, respectively, were collected. For the short signal, we used a phase-fitting technique and found P1= 6 120 ± 2 s. For the 64
ks period, the number of sampled cycles, approximately 3, is too small for the phase fitting. We therefore binned the light curve to 6120 s, so to avoid beat-frequency signals produced by the shorter periodicity, and fit a sinusoidal function to it. The fit has a χ2
of 54 for 30 degrees of freedom (dof) and we derived the period P2= 64.2 ± 0.5 ks. The 0.5–10 keV background-subtracted light
curves folded on our best periods are shown in Fig.3. We measured the following pulsed fractions: 100 ± 4 per cent (P1 = 6.1 ks;
this value is to be regarded as a lower limit) and 70± 4 per cent (P2= 64.2 ks). Within the statistical uncertainties, the shape and the
pulsed fraction of both signals are energy independent. In the soft (<2 keV) and hard (>2 keV) bands, we measured pulsed fractions of 96± 5 and 106 ± 5 per cent for the 6.1 ks period, and 69 ± 5 and 64± 7 per cent for the 64.2 ks period.
The 6.1 ks signal is significantly detected also in the 2001 XMM– Newton/EPIC data (∼700 net counts between the three EPIC cam-eras), while the observation is too short for the 64.2 ks period (Fig.3). We measured the period P1= 6.04 ± 0.04 ks and a pulsed
fraction of 88± 12 per cent. CXO J141430 is detected with ∼170 net counts in the 2014 XMM–Newton pointing (pn plus MOS 2, in the MOS 1 the source fell in one of the failed CCDs). The short-period pulsations are also clear in that data set, but the low count statistics hampers a precise estimate of the period. Finally, CXO J141430 was in the field of view of Chandra also in the observations 355 (2000 January, 1.3 ks) and 356 (2000 March, 25 ks; Table1). In observation 355, CXO J141430 is detected with a dozen of pho-tons only, and a signal-to-noise ratio SNR 3: the short duration
and the very small number of photons preclude any analysis of the periodic signals. In the data set 356, the source is detected with about 90 photons (SNR > 9). The 6.1 ks signal can be clearly ob-served but, similarly than in the 2014 XMM–Newton observations, the uncertainty on the period is very large.
4.2 Spectral analysis
For the spectral analysis, we started from the long Chandra observa-tion 12823. The fits were performed between 0.6 and 6 keV because of the very low signal of CXO J141430 outside this range. We fit to the data a power-law model, a blackbody, and an optically thin ther-mal bremsstrahlung. The blackbody model yielded a reduced χ2
ν =
1.37 for 40 dof with clearly structured residuals; the derived temper-ature is kT= 0.81 ± 0.03 keV, while for the absorption there is only an upper limit of NH< 0.8 × 1022cm−2at 90 per cent confidence.
The observed flux was FX= 6.8+0.5−0.4× 10−14 erg cm−2s−1 (0.5–
10 keV). The power law and the bremsstrahlung gave somewhat bet-ter fits, χ2
ν = 1.27 and 1.24, respectively, and better residuals. The
parameters of the power-law fit (Fig.4, left) are NH= 0.36+0.14−0.13×
1022 cm−2, = 1.51+0.16
−0.15, and FX = (1.03 ± 0.09) × 10−13
erg cm−2s−1. For the bremsstrahlung, NH= 0.31+0.11−0.10× 1022cm−2,
kT = 13+12−5 keV, and FX= 0.97+0.09−0.10× 10−13erg cm−2s−1.
In the second Chandra observation, the flux was ≈30 per cent higher. The absorption was only poorly con-strained (<0.6 × 1022 cm−2 at 90 per cent confidence for the
power law and <0.5 × 1022cm−2for the bremsstrahlung), while
kT and were consistent with those measured in observation 12823. We thus decided to fit the two spectra simultaneously, with the normalizations free to vary and the other parameters tied up between the data sets. The results are summarized in Table2.
The 2001 XMM–Newton data flatly reject the blackbody model, with χ2
ν = 2.33 for 28 dof. The power law provides a good fit to the
Figure 4. Chandra/ACIS spectra and best-fitting power-law models (red solid line) for CXO J141430 and CXOU J141332 (as indicated in each panel) from observation 12823. Bottom panels: residuals in units of standard deviations.
Table 2. Spectral results of CXO J141430. Errors are at a 1σ confidence level for a single parameter of interest.
Model Data NHa kT Fluxb Unabsorbed fluxb χν2(dof)
(1022cm−2) (keV) (10−13erg cm−2s−1)
PHABS(POWERLAW) Chandra/12823 0.30+0.13−0.11 1.43+0.14−0.13 – 1.07± 0.08 1.21± 0.07 1.12 (52)
Chandra/12824 1.40+0.15−0.14 1.58+0.13−0.12
PHABS(BREMSSTRAHLUNG) Chandra/12823 0.27± 0.10 – 18+24−7 1.01+0.10−0.09 1.13± 0.08 1.10 (52)
Chandra/12824 1.32+0.16−0.15 1.47± 0.14
PHABS(POWERLAW) XMM/0111240101 <0.3c 1.10± 0.15 – 0.87± 0.08 0.89+0.07−0.06 1.02 (28) Notes.aThe abundances used are those of Wilms, Allen & McCray (2000); N
Hvalues≈30 per cent lower are derived with those by Anders & Grevesse (1989). The photoelectric absorption cross-sections are from Balucinska-Church & McCammon (1992).
bIn the 0.5–10 keV energy range.
cUpper limit at the 90 per cent confidence level.
data, with an observed flux similar to that of the first Chandra ob-servation. The bremsstrahlung fit was equally good. Its temperature, however, could not be constrained, as it always pegged to the highest allowed value, showing that in the XMM–Newton data its curvature is indistinguishable from that of a power law. For this reason, in Table2we give only the parameters derived from the power-law fit. The quality of the spectrum from the 2014 XMM–Newton data is too poor for a spectral analysis. We thus used the models in Table2to es-timate the flux of CXO J141430, obtaining FX= (4.8 ± 0.6) × 10−14
erg cm−2s−1for the power-law model and (3.4± 0.5) × 10−14 erg cm−2s−1for the bremsstrahlung. Similarly, for the 2000 Chan-dra observation 356 (∼90 counts in a front-illuminated ACIS-S CCD), we evaluated a flux FX= (1.1 ± 0.2) × 10−13erg cm−2s−1
for the power-law model and (8.0± 0.9) × 10−14 erg cm−2 s−1 for the bremsstrahlung. In the short pointing 355, where only a dozen of photons were detected in one of the ACIS-I CCDs,
we converted the count rate into a flux with PIMMS2 and found
FX= (2.0 ± 0.7) × 10−13 erg cm−2s−1for the power-law model
and (1.5± 0.5) × 10−13erg cm−2s−1for the bremsstrahlung.
5 T H E 1 . 8 H P U L S AT O R : C X O U J 1 4 1 3 3 2 5.1 Timing analysis
CXOU J141332 displays a periodicity at roughly 6.4 ks (Fig.5). This signal was detected at about 3.5σ confidence level in observa-tion 12823 (∼270 counts) and is present also in the shorter pointing
2We used the web version of
PIMMS (Portable, Interactive Multi-Mission Simulator) available athttp://heasarc.gsfc.nasa.gov/cgi-bin/Tools/ w3pimms/w3pimms.pl.
1118
P. Esposito et al.
Figure 5. Same as Fig.3, but for CXOU J141332. For displaying purposes, the Chandra data have been shifted in power by+50.
12824 (∼80 counts), though at a lower confidence level. By the phase-fitting analysis, we derived the value P= 6377 ± 4 s. In the corresponding 0.5–10 keV background-subtracted folded pro-file (Fig. 5), we measured a pulsed fraction of 56 ± 8 per cent. The modulation of CXOU J141332 changes in shape and pulsed fraction as a function of energy: the asymmetry in the profile be-comes more accentuated, while the pulsed fraction decreases as the energy increases: 78± 7 per cent in the soft range (<2 keV) and 40± 8 per cent in the hard range (>2 keV). Of all the other obser-vations in Table1, CXOU J141332 was detected only in the 2001 XMM–Newton data, where the harvest was∼440 source photons between the pn and MOS 1. We measured a period of 6.4± 0.1 ks and the pulsed fraction was 59± 10 per cent.
5.2 Spectral analysis
Given the paucity of photons (all in the 1–6 keV range), we per-formed the spectral analysis by fitting the same spectral models adopted for CXO J141430 to the two Chandra data sets 12823 and 12824 simultaneously, with the normalizations free to vary and the other parameters tied up. The blackbody model provides a poor fit to the data, with a χ2
ν = 1.68 for 13 dof, while for the
bremsstrahlung (χ2
ν = 1.31) the temperature could not be
con-strained. For this reason, in Table 3 we give only the parame-ters obtained from the power-law fit (see Fig. 4, right), which gave a good fit with a rather hard power law with photon index ≈ 0.8 and a best-fitting absorption NH= 0.2 × 1022cm−2(with
a 90 per cent upper limit of 0.9× 1022cm−2). The observed flux
was FX≈ 4 × 10−14erg cm−2s−1during the first observation and
≈6 × 10−14 erg cm−2s−1in the second; when one considers the
large uncertainties, the flux increase is however only marginally significant.
The 2001 XMM–Newton observation offers a better spectrum, covering with a few more photons the band 0.4–8 keV. The power-law fit (χ2
ν = 1.02 for 16 dof) yielded photon index and flux similar
to those derived with Chandra (see Table 3) and made it possi-ble to constrain better the absorbing column. This was measured at NH= 0.13+0.09−0.07× 1022cm−2. In the 2014 XMM–Newton
obser-vation, CXOU J141332 was not detected. The 3σ upper limit on its observed flux, derived withPIMMSassuming the XMM–Newton power-law model in Table3, was 2× 10−14 erg cm−2s−1in the 0.5–10 keV band in the MOS data (in the pn, the position of the source occurred in proximity to streaks of out-of-time events due to the nucleus of the CG).
The position of CXOU J141332 was also imaged in the Chan-dra observations 2454, 355, 356, 10850, 10872, and 10873 (see Table1); the source was never detected and for each data set we derived in a like manner the following upper limits (for the grating observations, we considered only the zero-order data): 1.1× 10−13 erg cm−2 s−1(obs. 355, ACIS-I), 10−14 erg cm−2s−1 (obs. 356, ACIS-S), 7× 10−14erg cm−2s−1(obs. 2454, ACIS-S), 5× 10−14 erg cm−2s−1(obs. 10873, ACIS-S), 8× 10−14erg cm−2s−1(obs. 10850, ACIS-S), and 6× 10−14erg cm−2s−1(obs. 10872, ACIS-S).
6 A S T R O M E T RY A N D O P T I C A L O B S E RVAT I O N S O F C X O J 1 4 1 4 3 0 A N D C X O U J 1 4 1 3 3 2
6.1 X-ray astrometry
In order to improve the absolute astrometry of the Chandra data to search for optical counterparts to CXO J141430 and CXOU J141332, we cross-correlated the X-ray source list obtained usingWAVEDETECTwith sources in the Two-Micron All-Sky Survey
(2MASS; Skrutskie et al.2006) catalogue, which has an astromet-ric accuracy better than 0.2 arcsec. We found 17 2MASS point sources coincident within 0.4 arcsec from an X-ray source and used them to register the Chandra images on the accurate 2MASS ref-erence frame by fitting a transformation matrix which includes a rotation, scale factor, and translation. We note that the Chandra– 2MASS superposition did not require a significant transformation: the corrections are of the same order of the residuals (<0.15 arc-sec). The resulting positions (J2000.0) of the new CATS @ BAR pulsators are RA= 14h14m30.s1 (±0.s45) and Dec. = −65◦1623.3
(±0.30) for CXO J141430, and RA= 14h13m
32.s9 (±0.s30) and
Dec. = −65◦1756.5 (±0.25) for CXOU J141332, where the 1σ
−1.3 −1.0
PHABS(POWERLAW) XMM/0111240101 0.13+0.09−0.07 0.93± 0.13 – 5.4± 0.5 5.6+0.5−0.4 0.82 (16) Notes.aThe abundances used are those of Wilms et al. (2000); NHvalues≈30 per cent lower are derived with those by Anders & Grevesse (1989). The photoelectric absorption cross-sections are from Balucinska-Church & McCammon (1992).
bIn the 0.5–10 keV energy range.
cUpper limit at the 90 per cent confidence level.
Table 4. VST/OmegaCAM images of regions around CXO J141430 and CXOU J141332.
MJD Exp. Filter Archive name (s) 56442.149 25 r OMEGA.2013-05-30T03:34:12.715 56442.156 25 r OMEGA.2013-05-30T03:44:47.517 56460.067 25 r OMEGA.2013-06-17T01:07:16.243 56460.054 25 r OMEGA.2013-06-17T01:18:02.485 56487.025 25 r OMEGA.2013-07-14T00:36:46.111 56487.014 40 r OMEGA.2013-07-14T00:20:12.159 56488.033 25 r OMEGA.2013-07-15T00:47:30.554
uncertainties combine the Chandra localization accuracy, the resid-uals of the Chandra–2MASS frame superposition, and the 2MASS absolute astrometric accuracy.
6.2 VST data
Optical images of the regions around CXO J141430 and CXOU J141332 were retrieved from the ESO Science Archive Fa-cility. The observations were originally obtained with the 2.6 m VST located at Paranal Observatory using the OmegaCAM instru-ment (Kuijken2011), as part of the VST Photometric Hα Survey of the Southern Galactic Plane and Bulge (Drew et al.2014). Table4
lists the details of the images examined.
We aligned and stacked the images using the Image Reduction and Analysis Facility (IRAF) software and theIMALIGNandIMCOMBINE
packages following standard procedures. A world coordinate system was then applied using theIMWCSutility3and the Third US Naval
Observatory CCD Astrograph Catalog (Zacharias et al.2010). The mean error in the world coordinate system is 0.35 arcsec (3σ ) using 22 local stars. We performed aperture photometry on the nearby fields using theIRAFpackageAPPHOTand estimate a limiting
magnitude depth of r≈ 22.5 mag. No sources are observed within 2 arcsec of the coordinates of CXO J141430 and CXOU J141332.
7 D I S C U S S I O N I : T H E N AT U R E O F C X O J 1 4 1 4 3 0 A N D C X O U J 1 4 1 3 3 2
With a new X-ray pulsator discovered by CATS @ BAR ev-ery∼2200 light curves, the detection of two previously unknown pulsating sources in the Circinus data set had a formal probability of about 0.04 per cent. While this does not qualify as a statisti-cal anomaly, the 2010 Chandra observations have certainly been bountiful for the CATS @ BAR project.
3Seehttp://tdc-www.harvard.edu/wcstools/.
7.1 CXO J141430
In the case of CXO J141430, which is displaced by ∼2 arcmin from the extreme edge of CG (Fig.1), the two periodicities nail it down as an intermediate polar (IP) with spin period of Pspin= 6.1
ks (1.7 h) and orbital period Porb = 64.2 ks (17.8 h). IPs, also
known as DQ Herculis stars, and polars (AM Herculis stars) are the two main subclasses of magnetic cataclysmic variable stars (CVs). CVs are close binaries hosting a white dwarf (WD) accreting from a late-type Roche lobe filling companion, either a main-sequence or a sub-giant star. IPs are characterized by asynchronous rotation (Pspin < Porb), while polars are phase-locked (Pspin Porb) and
generally display strong circular polarization (whence the name) at optical and near-infrared wavelengths (see Patterson1994; Warner
2003; Smith2006for reviews). Although it is still matter of debate, these differences are generally interpreted as due to a magnetic field in IPs which is weaker than that typically measured for polars (B 107G).
The orbital period of CXO J141430 locates the system above the 2–3 h so-called orbital period gap, where most of the IPs are found. Also the spin-to-orbit period ratio of∼0.095 is typical of an IP (it is generally in the range 0.25–0.01, with most systems around 0.1).4In
IPs, the accreted material generally passes through a disc and is then channelled on to the magnetic polar regions of the WD (at variance with polars, where the magnetic field inhibits the formation of the disc). There, a shock develops and the hot gas cools while it settles on to the WD surface emitting X-rays via thermal bremsstrahlung and cyclotron radiation (Aizu1973). Because of their strong mag-netic field, in polars the cooling takes place mainly via cyclotron, whereas IPs are expected to show bremsstrahlung-dominated emis-sion. While the relatively poor statistical quality of the available spectra precluded a good characterization of the X-ray emission of CXO J141430, the results of the spectral analysis are consistent with this picture (Table 2). Also the large-amplitude modulation at the orbital period is rather common in IPs (e.g. Parker, Norton & Mukai2005), and the∼100 per cent pulsed fraction hints at a high-inclination system. We finally note that the X-ray luminosity of CXO J141430 in the deep Chandra and XMM–Newton observa-tions was LX≈ 2 × 1031d12erg s−1(with a≈50 per cent variability,
see Section 4), where d1is the distance in units of 1 kpc. Typical
values for IPs, 1032–1034erg s−1(Sazonov et al.2006), suggest that
CXO J141430 is either on the lower side of the luminosity distribu-tion or the distance to the source is substantially larger than∼1 kpc. For an IP with orbital period of 17.8 h, a K5V star would be a likely companion (e.g. Smith & Dhillon1998). Using a value of NH= 0.3 × 1022 cm−2 derived from our model fit to the X-ray
4See for example the catalogue available at the Intermediate Polar Home Page,http://asd.gsfc.nasa.gov/Koji.Mukai/iphome/catalog/members.html.
1120
P. Esposito et al.
spectra, and a conversion of NH/AVof 1.79× 1021 cm−2mag−1
(Predehl & Schmitt1995), we obtain AV= 1.7 mag. Assuming an
absolute magnitude Mr≈ 7.1 mag (Bilir et al.2008),5the limiting
magnitude mr ≈ 22.5 suggests d > 5 kpc.
7.2 CXOU J141332
The nature of the fainter CXOU J141332 is less obvious. The source is located inside the 90 per cent total B light contour of the CG. So, the first question that needs to be addressed is whether it is a Galactic or an extragalactic source. Based on the cumulative Galactic X-ray source density versus flux distribution (log N–log S) from the Chandra Multi-wavelength Plane survey (ChaMPlane; van den Berg et al. 2012), we estimated the prob-ability of a foreground Galactic object of that flux within this area to be≈11 per cent. Moreover, its absorbing column measured with XMM–Newton (NH= 1.3+0.9−0.7× 1021cm−2; Table3) is much
lower than the total Galactic value of∼6 × 1021cm−2(Dickey &
Lockman1990; Kalberla et al.2005). CXOU J141332 is therefore most likely a Galactic source.
The hard X-ray spectrum, a power law with photon index ∼ 0.8–0.9, and the modulation at 6.4 ks (1.8 h) point to a bi-nary system consisting of a compact star accreting from a low-mass companion, where the 1.8 h period likely traces the orbital motion. The period is in fact too long to be the spin of a typical neutron star (NS; with very few possible exceptions; Mattana et al.2006; Esposito et al. 2011). The low flux and – chiefly – the smooth and ∼60 per cent pulsed-fraction modulation favour a magnetic CV nature also for CXOU J141332 (the period is also too short for the orbit of a standard high-mass X-ray binary). Indeed, CVs are the most abundant population of Galactic compact interacting binaries, and also the most frequent new pulsating sources in the CATS @ BAR sample (Israel et al., in preparation). In particular, since no second periodicity was detected (and lacking any informa-tion about optical polarizainforma-tion), the source could be a polar. Polars are generally found at short orbital periods, most of them below the 2–3 h orbital gap, and CXOU J141332 would lie in the peak of their period distribution (e.g. Ritter & Kolb2003). The profile of the folded light curve (Fig.5) and its variability as function of energy may indicate a two-pole system. The luminosity during the long 2010 Chandra observations was LX≈ (5–8) × 1030d12erg s−1,
with the upper limits from the other observations implying a vari-ability of≈50 per cent or larger. For distances of the order of a few kpc, this is in good agreement with typical values for polars (LX<
1032erg s−1; Sazonov et al.2006).
For a polar with orbital period of 1.8 h, the companion is likely an M5V star, which has an absolute magnitude of Mr ≈
12.5 mag (Bochanski, Hawley & West 2011). The value of NH = 0.13 × 1022 cm−2 derived from our model fit to the
X-ray spectra implies an AV = 0.7 mag. The limiting magnitude
mr≈ 22.5 suggests d 0.7 kpc. In the IP hypothesis, assuming
for CXOU J141332 a K5V as for CXO J141430, the non-detection would indicate a distance larger than∼8 kpc.
8 T H E C O N T R OV E R S I A L S O U R C E C G X - 1
CG X-1 (CXOU J141312.3–652013), about 15 arcsec north-east of the Circinus’ nucleus, had been known for long to be a bright and
5Uncertainty in the assumed absolute magnitudes of the companion star may be as large as 2 mag. The same holds for CXOU J141332, see below.
variable (possibly periodic) X-ray source. Using high-quality light curves collected with Chandra, Bauer et al. (2001) discovered a strong modulation at a period of∼27 ks. The measured X-ray flux was 9× 10−13erg cm−2s−1(0.5–10 keV), and deep Hubble Space Telescope (HST) observations did not detect any optical counterpart to CG X-1, with a limit mF606W > 25.3. They observed that the
source might be either a black hole (BH) binary in the CG radiating at∼4 × 1039erg s−1(and hence qualifying as an ULX) or a Galactic
CV of the polar type with a particularly long period (in both cases, the 27 ks modulation would reflect the orbital period of the system). In the polar hypothesis, for an M2V to M6V companion star, the HST limit puts the source at a distance larger than 1.2 kpc, implying a luminosity of at least 3× 1032erg s−1, a rather extreme value for a
polar. Bauer et al. (2001) also noticed that the association of CG X-1 with the CG is convincing: based mainly on the results from the ASCA Galactic plane survey (Sugizaki et al.2001), they evaluated that the possibility of a foreground or background X-ray source is 0.06 per cent. Moreover, Smith & Wilson (2001) noticed that the absorption towards CG X-1 (NH> 1022cm−2) is much larger than
the total Galactic column (NH∼ 6 × 1021cm−2), further supporting
the association with the CG. Overall, Bauer et al. (2001), Smith & Wilson (2001), and Bianchi et al. (2002) favoured a very bright extragalactic BH binary, harbouring a BH possibly in excess of 50 M.
Despite recognizing the robustness of the association, Weisskopf et al. (2004) were more open towards the possibility of a foreground polar. They observed that while the period and the luminosity would be somewhat atypical, CG X-1 would be neither the longest period nor the brightest known polar. On the other hand, they argued that if CG X-1 belonged to the CG, because of the short orbital period it should be a BH low-mass X-ray binary (LMXB) with a1 M companion. In this case, the huge X-ray luminosity of the BH would drive the star out of thermal equilibrium and evaporate it within∼103yr. They regard as very unlikely the possibility that a
system this short lived could be observed.
Weisskopf et al. (2004) revised the period of CG X-1 at 26.25± 0.15 ks and noticed a possible optical counterpart with mF606W= 23.5. However, they did not estimate the significance of
the association or of the source, and gave instead a limiting magni-tude of 24.3. Ptak et al. (2006) confirmed the limit by Bauer et al. (2001), while a recent work by Gladstone et al. (2013) proposed a counterpart with mV= 24 ± 6 [presumably, the same excess/source
detected by Weisskopf et al. (2004)]. In the literature, CG X-1 is generally considered to be an ULX in the CG (e.g. Bianchi et al.
2002; Swartz et al.2004; Liu & Mirabel2005; Ptak et al.2006; Berghea et al.2008; Gladstone et al.2013).
9 T H E N E W CHANDRA DATA O F C G X - 1 : A N A LY S I S A N D R E S U LT S
Since the CG has been observed many times in X-rays, in particular with Chandra, a wealth of data exist for CG X-1. Detailed studies of CG X-1 with ROSAT, BeppoSAX, XMM–Newton, and Chandra were presented in the aforementioned works by Smith & Wilson (2001), Bauer et al. (2001), Bianchi et al. (2002), and Weisskopf et al. (2004) (but see also Matt et al.1996; Guainazzi et al.1999; Sambruna et al.2001; Massaro et al.2006; Bauer et al.2008; Yang et al.2009; Shu, Yaqoob & Wang2011; Walton et al.2013; Ar´evalo et al.2014). A systematic analysis of all the available data is beyond the scope of this paper, and here we will only present results from the analysis of observations 12823/4, which represent the deepest
(0.06 per cent) estimated by Bauer et al. (2001). We actually believe that the probability of an interloper is substantially lower (approximately two times smaller), considering that the flux reached by the source in subsequent observations is∼5 times higher (Weis-skopf et al.2004) and the fact that a background AGN can be excluded by the phenomenology of CG X-1.
For the spectral and timing analysis, we extracted the source counts within a 1.5 arcsec radius, while for the background we used an annulus with radii of 3 and 5 arcsec (see Fig.1and Sec-tion 2 for more details). The 0.3–8 keV source net count rate was (6.90± 0.07) × 10−2 counts s−1 in observation 12823 and (6.1± 0.1) × 10−2counts s−1in observation 12824. These rates are high enough to cause pileup in the ACIS detector (as we checked with a pileup map). For the spectral analysis, the pileup was dealt with by using the pileup model by Davis (2001). This procedure involves some uncertainty, because the pileup fraction in CG X-1 is strongly dependent on the orbital phase. However, none of our re-sults crucially depends on the exact value of the parameters derived from the spectral fitting.
About six 26 ks cycles were recorded in obs. 12823 and two in obs. 12824. There is a moderate pulse-to-pulse variability, both in shape and in the flux at maximum (≈30 per cent). Also, the average count rate was≈15 per cent lower during the second observation. To measure the period, we fit a sinusoidal function to the light curve. We obtained PCG X−1= 26.1 ± 0.1 ks. The corresponding folded
profile in different energy bands is shown in Fig.6. The profile is asymmetric and the modulation is large (but the count rate is non-zero also at minimum). The pulsed fraction, defined as6(M−
m)/(M + m), where M is the maximum count rate and m the mini-mum, is 91.6± 1.5 per cent in the whole band, 89.2 ± 2.5 per cent in the 0.3–2 keV band, and 97.1± 1.2 per cent in the 2–8 keV band. A conspicuous spectral softening around minimum is evident from the ratio of the hard to soft counts along the cycle (Fig.6).
For the spectral analysis, we fit three simple models to the data: a power law, a multicolour disc (MCD; Mitsuda et al.1984; Mak-ishima et al.2000), and an optically thin thermal bremsstrahlung, all corrected for the interstellar absorption. While an MCD with kT 1.3–1.4 keV gives the lowest χ2, all models provide an acceptable
fit to the data (Table5). All fits confirm that NHtowards CG X-1
is substantially larger than the total Galactic absorbing column in that direction. Weisskopf et al. (2004) reported a possible feature, probably a blend of Fe lines, in the 2001 XMM–Newton spectrum of CG X-1. No line is required to fit the Chandra data (see Fig.7for the longest observation); the 3σ upper limit on the equivalent width of any line with central energy between 6 and 7 keV is 0.18 keV in observation 12823. This limit is formally compatible with the equivalent width of 0.23± 0.06 keV derived by Weisskopf et al. (2004). However, since there is no trace of such feature in the ACIS data, it is possible that, as noticed also by Weisskopf et al. (2004), the feature observed with XMM–Newton is due to residual
contam-6Here we adopted a different definition of the pulsed fraction than used before, because the value inferred from a sinusoidal fit would misrepresent the amplitude of the modulation of the ‘sawtooth’ profile of CG X-1.
Figure 6. Background-subtracted folded profile of CG X-1 (observation 12823/4) in different energy bands. The hardness ratio between the hard and soft bands is also plotted at the bottom.
ination from the emission of the nuclear region of the CG, which has very strong lines at 6.4 and 7 keV.
We also extracted pulse-resolved spectra from the soft (phase between 0.55 and 0.9 in Fig.6) and hard (all other phases) parts of the hardness ratio. The data were prepared withDMTCALC, and the
spectra of the two data sets from the same phase bins were combined usingCOMBINE_SPECTRA, which also averaged the response matrices.
Although the few counts in the soft spectrum (∼700 versus more than 12 200 in the hard spectrum, after combining the two obser-vations) preclude a detailed comparison, we found that the spectral variation can be described equally well by either a decrease in the absorption during the softening [NH= (1.00 ± 0.03) × 1022cm−2
in the hard phase and (0.24 ± 0.07) × 1022 cm−2 in the soft
phase for the MCD model; (1.45 ± 0.04) × 1022 cm−2 in the
hard phase and (0.67± 0.08) × 1022 cm−2in the soft phase for
the power-law model; (1.29 ± 0.03) × 1022 cm−2 in the hard
phase and (0.51± 0.07) × 1022 cm−2 in the soft phase for the
bremsstrahlung model] or a change in the pivotal parameter of the model (kT= 1.82 ± 0.05 keV in the hard phase and 0.95 ± 0.06 keV in the soft phase for the MCD; = 1.70 ± 0.03 in the hard phase and 2.54± 0.11 in the soft phase for the power law; kT = 9.4 ± 0.8 keV
1122
P. Esposito et al.
Table 5. Spectral results of CG X-1. Errors are at a 1σ confidence level for a single parameter of interest.
Model Obs. ID NHa kT Fluxb Unabsorbed fluxb χν2(dof)
(1022cm−2) (keV) (10−12erg cm−2s−1) PHABS(DISKBB) 12823 1.06± 0.04 – 1.36± 0.07 0.83+0.13−0.12 1.17± 0.17 0.97 (271) PHABS(POWERLAW) 12823 1.38± 0.05 1.72−0.05+0.06 – 1.23+0.10−0.09 1.84+0.12−0.10 1.04 (271) PHABS(BREMSSTRAHLUNG) 12823 1.28± 0.04 – 6.5+1.0−0.8 1.02+0.04−0.05 1.50+0.05−0.06 1.00 (271) PHABS(DISKBB) 12824 1.03+0.08−0.07 – 1.33+0.14−0.10 0.68+0.21−0.16 0.85+0.29−0.22 0.98 (92) PHABS(POWERLAW) 12824 1.38+0.11−0.10 1.77+0.12−0.10 – 1.00+0.15−0.14 1.52+0.18−0.17 1.11 (92) PHABS(BREMSSTRAHLUNG) 12824 1.27± 0.08 – 5.6+1.7−1.2 0.83± 0.07 1.25± 0.09 1.05 (92) Notes.aThe abundances used are those of Wilms et al. (2000); NHvalues≈30 per cent lower are derived with those by Anders & Grevesse (1989). The photoelectric absorption cross-sections are from Balucinska-Church & McCammon (1992).
bIn the 0.5–10 keV energy range.
Figure 7. Chandra/ACIS spectrum and best-fitting MCD model (red solid line) for CG X-1 from observation 12823. Bottom panel: the residuals of the fit in units of standard deviations.
in the hard phase and 2.4 ± 0.3 keV in the soft phase for the bremsstrahlung).
1 0 D I S C U S S I O N I I : I S C G X - 1 A
W O L F – R AY E T / B H B I N A RY I N T H E C G ?
While we cannot completely exclude the chance superposition of a Galactic polar in the direction of the inner part of the CG, we regard the association of the CG X-1 to the CG as rather compelling. In the following, we will discuss the nature of the source in this framework.
Weisskopf et al. (2004) disfavoured the possibility of a BH LMXB because such a system would be rather short lived, and thus unlikely to be observed. This argument is not conclusive when dealing with an individual source (one can be lucky enough to observe a rare system!), but we too deem an LMXB as unlikely. This is because of its orbital profile (Fig.6). In fact, most LMXBs show no or very low amplitude modulation on their orbital period. In those which display orbital modulation, dipping and/or eclipsing systems, the morphology of the profile is very different. In dipping systems, the dips are produced by absorption of X-rays due to accreting matter located in the bulge at the outer edge of the accretion disc (e.g.
D´ıaz Trigo et al.2006); the X-ray minima tend to be rather sharp and to show a harder-than-average emission, which is the opposite of what is observed in CG X-1 (Fig. 6, bottom panel). LMXB eclipses display sharp and abrupt ingresses and egresses, due to the small size of the X-ray-emitting regions. Even the smoothly modulated accretion disc corona sources (systems viewed nearly edge-on where the outer edge of the dense disc modulates the X-rays from the central source scattered into the line of sight by an extended ionized corona; Mason & Cordova1982; White & Holt1982; Somero et al.2012) have different profiles. Moreover, LMXBs containing BHs are usually transient X-ray sources, with outburst durations of the order of weeks to months (e.g. Remillard & McClintock2006), while CG X-1 is variable but persistent.
On the other hand, the light curve and the folded profile of CG 1 bear a strong resemblance to those observed in high-mass X-ray binaries (HMXBs) with a Wolf–Rayet (WR) star companion: Cyg X-3 in the Milky Way (P = 4.8 h; Zdziarski et al. 2012), IC 10 X-1 in IC 10 (P= 35 h; Prestwich et al.2007), NGC 300 X-1 in NGC 300 (P= 33 h; Carpano et al.2007b), and the candi-dates CXOU J123030.3+413853 in NGC 4490 (CXOU J123030; P= 6.4 h; Esposito et al.2013c) and CXOU J004732.0−251722.1 in NGC 253 (with a candidate periodicity P∼ 14–15 h; Maccarone
studied source of this kind is Cyg X-3, where the orbital modula-tion has been ascribed to attenuamodula-tion by electron scattering in the strong WR wind (Hertz, Joss & Rappaport1978). A further sig-nificant contribution to the scattering medium is thought to be a bulge of ionized matter formed by the collision of the stellar wind with the outer accretion disc (Zdziarski, Misra & Gierli´nski2010). Alternative explanations involve orbital modulation by absorbers in different phases (hot/ionized and cold/clumpy, triggered by the BH jet bow shock; Vilhu & Hannikainen2013), or the presence of the accretion wake, a large-scale asymmetry around the compact object (Okazaki & Russell2014). The orbital profile of CG X-1 is even more asymmetric than in Cyg X-3, probably due to rather extreme properties of the absorbing/scattering medium or because of a higher inclination of the system. The variation of the hardness ratio along the orbit (Fig.6), showing a clear softening during the X-ray minimum, could be due to direct X-rays almost completely blocked by dense matter (probably the innermost regions of the WR wind when the X-ray source is at the superior conjunction). The softer X-rays observed could be due to down-scattering into the line of sight of central X-rays by cold and less dense material located farther away from the WR star (similarly to what usually observed during eclipses in some HMXBs; Haberl1991).
Weisskopf et al. (2004) mentioned for CG X-1 the possibility of a WR or, more in general, a naked He donor, but did not discuss it in detail owing to the scarcity of information of such systems. In particular, they observed that for an∼2 M companion, their argu-ment against LMXB systems was still relevant. However, thanks to their compactness, even more massive WR stars can comfortably fit in the orbit of a system like CG X-1 (the stellar radius of a 20 M WR star is <2 R; e.g. Langer1989; Schaerer & Maeder1992). Also the other main observational properties of CG X-1 fit well in the scenario of a WR–BH HMXB in the CG.8
If one considers the reddening towards the CG (4 mag) and its distance module (28.1 mag, from NED), the limit on the optical counterpart by Bauer et al. (2001) and Ptak et al. (2006) implies MV> −6.8. This value is compatible with a WR star, for which MV
is typically in the range from−2.5 to −7 (e.g. Massey2003). The X-ray luminosity of CG X-1 is variable by a factor of≈10 (Bianchi et al.2002; Weisskopf et al.2004). The highest flux re-ported in the literature (5.2× 10−12erg cm−2s−1for a power-law fit or 5× 10−12erg cm−2s−1for an MCD fit, in the 0.5–8 keV band; Weisskopf et al.2004) would imply, for a distance of 4.2 Mpc, a 0.5– 10 keV luminosity of LX= (1.5–2) × 1040erg s−1. If the system is
Eddington-limited, the lower limit on the mass of the accreting BH is MBH 75 M for an He or C/O donor. For the system to shine
in X-rays, the velocity of the WR star wind has to be slow enough to allow the formation of an accretion disc. This condition
corre-7Apart from these objects, the only other known WR HMXB is ULX-1 in M101 (Liu et al.2013, see also Section 10.1). A period of 8.2 d was inferred from radial velocities of optical emission lines, but no X-ray light curves of good quality are available for this source.
8Even bright (L
X> 1040erg s−1) ULXs may contain an NS (King2009; Bachetti et al.2014), although such systems are probably not the majority (Fragos et al.2015). Here, we will not discuss this possibility.
LX≈ η ˙ Mwc2G2MBH2 a2(v2 orb+ v2w)2 , (1)
where η is the efficiency, ˙Mw is the wind mass-loss rate, a is
the orbital separation, vorb is the orbital velocity, and vw is the
wind velocity at the BH orbit. Assuming ˙Mw= 10−5 M yr−1
and vw = 1000 km s−1 for the WR star (e.g. Crowther 2007),
a= 5.8 × 1011cm (for a 10 M
companion), MBH= 75 M, and
the formation of a disc with η = 0.1, the corresponding luminosity is LX 2 × 1040erg s−1. More in general, for MBH> 10 M and
all the other things being equal, one finds LX 3 × 1039erg s−1.
In case of Roche lobe overflow, even higher X-ray luminosity could be achieved. However, we note that if CG X-1 is indeed a WR–BH binary, the WR star is probably not filling its Roche lobe [unless it is very massive; see for example the discussion of the case of Cyg X-3, where the orbital period is much shorter, in Szostek & Zdziarski (2008)]. An X-ray luminosity of∼2 × 1040erg s−1can
be therefore accounted for. We finally notice that, although we do not regard the question as crucial, the problem of the lifetime of the system discussed by Weisskopf et al. (2004) would be significantly attenuated, since the WR phase of a massive O-type star is thought to last a few×105yr (Meynet & Maeder2005).
10.1 Statistics, environment, and WR–BH binaries as ULXs
All known WR HMXBs have been mentioned in the previous sec-tion. Three have been established so far as certain WR–BH systems: IC 10 X-1, NGC 300 X-1, and M101 ULX-1. For the fourth WR– compact object binary, Cyg X-3 in our Galaxy, it is still debated whether the compact object is a BH or an NS. There are how-ever show-everal pieces of evidence (radio, infrared, and X-ray emission properties) that point to a 2–5 M BH, as suggested also by evolu-tionary models (e.g. Lommen et al.2005; Szostek & Zdziarski2008; Szostek, Zdziarski & McCollough2008; Shrader, Titarchuk & Sha-poshnikov 2010; Zdziarski, Mikołajewska & Belczy´nski 2013). Apart from CG X-1, discussed in this paper, two additional WR–BH binary candidates were found in the last two years: CXOU J123030 in NGC 4490 and CXOU J004732.0−251722.1 in NGC 253.
We expect WR–BH binaries to be associated with star-forming regions, since WR are young stars with massive progenitors (with zero-age main-sequence mass25 M). Table6shows the star formation rate (SFR) of the host galaxies of WR–BH binaries and binary candidates. The average SFR is∼2 M yr−1, which is quite high for nearby late-type galaxies. All metallicities listed in Table6
are sub-solar. Recent N-body and population synthesis simulations of young star clusters (Mapelli et al.2013; Mapelli & Zampieri
2014) suggest that≈2 per cent of all HMXBs powered by BHs in star-forming regions are BH–WR binaries, independent of star cluster metallicity.
The average luminosity of CG X-1 in the Chandra observations presented here makes it a bona fide ULX. The maximum lumi-nosity reached by the source, following Weisskopf et al. (2004), is LX= (1.5–2) × 1040erg s−1. At such flux levels, in high counting
statistics XMM–Newton spectra, a persistent ULX typically shows the hallmark of the ultraluminous state (Gladstone, Roberts & Done
2009), with the presence of two thermal components, one of which
1124
P. Esposito et al.
Table 6. Properties of observed WR–BH binaries and candidates (denoted by stars), and of their host galaxies. Host galaxy Source Period BH massa WR massa SFRb Zb t
GWc (h) (M) (M) (Myr−1) (Z) (Gyr) IC 10 X-1 34.9 33 35 0.07 0.22 1.4 NGC 300 X-1 32.8 20 26 0.14 0.19 1.7 NGC 4490 CXOU J123030.3+413853* 6.4 – – 4.5 0.23 0.038 NGC 253 CXOU J004732.0−251722.1* 14.5 – – 4.0 0.24 0.33 Circinus CG X-1* 7.2 – – 1.5 0.10 0.052 M101 ULX-1 196.8 20 19 3.1 0.17 200
Milky Way Cyg X-3 4.8 3 7 0.25 0.31 0.051
aFor the BH and WR masses, we list only the fiducial values that we use to derive the merger rates R in equation (2). Most of these masses are very uncertain, as discussed in Prestwich et al. (2007), Silverman & Filippenko (2008), Carpano et al. (2007a,b), Crowther et al. (2010), Esposito et al. (2013c), Maccarone et al. (2014), Liu et al. (2013), and Shrader et al. (2010).
bThe values of SFR and metallicity of the host galaxy come from the compilation of Mapelli et al. (2010a). The metallicity
Z refers to the value at 0.7R25, where R25is the Holmberg radius of the galaxy, for NGC 253, NGC 300, NGC 4490, M101, and for the Milky Way, while it is the total metallicity for IC 10 and Circinus. We assume Z= 0.02. cThe parameter t
GWis the time-scale for the binary to coalesce, under the assumptions discussed in the main text.
centred at rather soft energies (∼0.1–0.2 keV) and the other pro-ducing a shallow but significant rollover around 3–5 keV. Such a spectrum is usually interpreted as the imprint of a super-Eddington accretion regime (e.g. Middleton, Sutton & Roberts2011).
As discussed in Section 8, the X-ray spectrum of the Chan-dra observations considered here is satisfactorily fit with a single-component model. The lack of evidence of multiple single-components may be real, but may also be caused by the comparatively low statistics (the flux is∼5 times lower than at maximum) and/or by the fact that Chandra has a lower sensitivity than XMM–Newton be-low∼0.5 keV, where the soft component peaks. If the spectrum is intrinsically single component, this may suggest that CG X-1 does not enter into the ultraluminous state and hence does not accrete above Eddington.
The latter possibility is consistent with the scenario discussed above in which accretion proceeds through a wind and the accretion rate does not exceed the Eddington limit. For wind accretion from a compact WR star, the very formation of a standard accretion disc is uncertain and the accretion efficiency can in general be smaller than that of a standard disc (e.g. Frank, King & Raine2002). A scenario of rather efficient accretion from a wind of a massive WR star on to a BH of a few tens M has been invoked also for M101 ULX-1. This source is a WR–BH ULX system with a dynamical mass measurement and a cool disc X-ray spectrum at maximum (Liu et al.2013), and it has a significantly larger orbital period (∼8 d,
see Table6) and smaller luminosity (∼3 × 1039erg s−1) than CG
X-1. Assuming a similar scenario also for CG X-1, simple BH mass estimates based on Eddington-limited accretion from an He–WR star (as those reported above) give MBH 70 M for the observed
maximum luminosity. Such a massive BH would populate the high-mass tail of the distribution of BHs formed through direct collapse of a massive star in a low-metallicity environment (Mapelli, Colpi & Zampieri2009; Zampieri & Roberts2009; Belczynski et al.2010), a scenario indeed consistent with the metallicity inferred for the CG (see Table6). In the same hypothesis, an independent limit can be obtained from the normalization of the disc component at maximum luminosity (Chandra Obs. ID 365; Weisskopf et al.2004), giving MBH 8[(d/4.2 Mpc)/
√
cos i] M (i is the inclination angle of the disc; e.g. Lorenzin & Zampieri2009).
On the other hand, the lack of a high counting statistics spec-trum at maximum luminosity prevents us from reaching a robust conclusion. We thus briefly consider also the possibility that the
system is accreting from a particularly massive and big WR com-panion via Roche lobe overflow, for which the mass transfer rate is expected to significantly exceed the Eddington limit (e.g. Lommen et al.2005). In this assumption, the maximum observed luminosity of CG X-1 would place it in the populated part of the ULX lumi-nosity distribution (e.g. Swartz et al.2011), where the observed flux can be produced by moderately beamed, super-Eddington emission from accretion on to a BH of a few tens M (or even a canonical stellar-mass BH).
It is interesting to compare the properties of the two candidate WR–BH ULX systems that we have tentatively identified, CG X-1 and CXOU J123030 in NGC 4490 (Esposito et al.2013c). While the orbital periods are similar (7.2 and 6.4 h, respectively), the X-ray luminosity is significantly different (CG X-1 being >10 times more luminous). Assuming a similar scenario of sub-Eddington accretion from a wind with the formation of a disc, the different luminosity could be mostly ascribed to the different BH mass, with CG X-1 being about∼10 times more massive than CXOU J123030. For non-extreme inclinations, this is consistent with the BH mass inferred from the normalization of the disc component of CXOU J123030, for which we obtained MBH 2.8[(d/8 Mpc)/
√
cos i] M (Es-posito et al.2013c). On the other hand, in case of non-standard super-Eddington accretion via Roche lobe overflow, the larger lu-minosity of CG X-1 may be caused by the larger accretion rate and we would then be witnessing the different X-ray outcome produced by similar BHs in very different accretion environments.
10.2 WR–BH binaries as precursors of BH–BH binaries
There is great uncertainty on the expected rate of BH–BH mergers in the frequency range that will be observed by Advanced LIGO and Virgo (∼10–104 Hz; Abadie et al.2010). In fact, while for
NS–NS binaries the expected merger rate can be derived from the properties of the observed NS–NS binaries (e.g. Kim, Kalogera & Lorimer2003) and from the rate of short gamma-ray bursts (e.g. Coward et al.2012; Fong et al.2012), no evidence has been found of BH–BH systems yet. WR–BH binaries can provide us with essential clues, since they are a possible precursor of BH–BH binaries (or BH–NS binaries), provided that the system is so tight that it remains bound when the WR star evolves into a compact remnant.
Here, we use the properties of all known WR–BH binaries and candidates to infer the BH–BH merger rate in the instrumental range
